Turn any song into MIDI, one instrument at a time
Upload the song and StemX separates it into six stems first, then transcribes each stem to its own MIDI file, so the bass notes don't get mixed up with the guitar. You also get one combined MIDI file with every part on its own track. Full songs use minutes; 30 are free each month.
We don't train models on your audio. Privacy policy
The free 60-second preview plays all six stems and has no MIDI. MIDI comes only with the full split, after you sign in. A full split uses one minute of your balance per minute of audio, and that minute covers all six stems and their MIDI. See plans and minutes.
How song to MIDI works
- 1
Upload the song
Drop an MP3, WAV, FLAC, M4A, OGG, AAC or AIFF file, up to 10 minutes and 120 MB.
- 2
Separate six stems
StemX splits the mix into vocals, drums, bass, guitar, piano and other. Hear the first 60 seconds of every stem free, before you sign in.
- 3
Transcribe each stem
Sign in and split the full song. Pitched stems are transcribed with YourMT3+ and the drums with ADTOF, one stem at a time.
- 4
Drag it into your DAW
Download a MIDI file per stem, or the combined file with every part on its own track, and drop it on your project.
Why it separates before it transcribes
Audio-to-MIDI on a full mix asks one model to hear every instrument at once. A bass note under a guitar chord can be read as part of the chord, the kick drum can turn into a low note, and the vocal melody gets tangled with whatever plays in the same range.
Separating first gives the transcriber one instrument at a time. The bass stem holds the bassline, so its MIDI file holds bass notes, and the guitar file holds the guitar part. Each part lands on its own track, ready to edit, re-voice or play with a different sound.
What you get for each stem
Vocals
The sung melody as single notes. The words aren't in the MIDI; transcribe the lyrics separately, free.
Drums
Hits sorted onto separate tracks: Kick, Snare, Toms, Hi-Hat and Crash & Ride.
Bass
The bassline. The bass plays one note at a time in a range of its own, which makes it one of the easier parts to transcribe.
Guitar and piano
Chords and melody lines together in one part per instrument. These are the hardest stems to transcribe; see the limits below.
Other
Everything else, such as strings, synths and horns, split into one track per instrument family.
Combined file
One MIDI file with every stem on its own track, so the whole arrangement opens at once.
What it gets wrong
Polyphonic piano and guitar are the hardest parts. Dense chords, fast strumming and the sustain pedal blur where one note ends and the next begins, so expect missed notes, extra notes and lengths to tidy up.
Drums come out as hits on the 5 kit tracks above. Open, closed and pedal hi-hat share the Hi-Hat track, and quiet ghost notes can be missed.
Vocals become the melody only. Slides, vibrato and bends are rounded to the nearest notes, so a smooth phrase comes out as steps.
The MIDI can only be as clean as the stem. Where a separated stem carries a little of another instrument, its MIDI can carry stray notes from it. A stem with nothing in it, such as piano on a song without piano, is marked as having no notes.
Treat the MIDI as a draft to edit in your DAW. StemX does not write sheet music or tabs.
Frequently asked questions
How do I convert a song to MIDI?
Upload the song at the top of this page. StemX splits it into six stems, and when you sign in and split the full song it writes a MIDI file for each stem plus one combined file.
Is MIDI included in the free preview?
No. The 60-second preview plays all six stems without MIDI. MIDI comes only with a full split.
How many minutes does it use?
One minute of balance per minute of audio, the same as any full split. That minute covers all six stems and every MIDI file. The Free plan includes 30 minutes a month; paid plans start at $4.99 for 300 minutes.
Can I convert an MP3 to MIDI?
Yes. You can upload MP3, WAV, FLAC, M4A, OGG, AAC and AIFF files, up to 10 minutes and 120 MB each.
Which instruments get MIDI?
All six stems: vocals, drums, bass, guitar, piano and other. Drums come as hits on Kick, Snare, Toms, Hi-Hat and Crash & Ride tracks, and the other stem is split by instrument family.
How accurate is the MIDI?
It depends on the part. Bass lines and single-note melodies come out closest to what was played; dense piano and strummed guitar need the most cleanup. We haven't published a MIDI accuracy figure yet, so try it on your own song first.
Can I get MIDI through the API?
Yes. Async full-split jobs return the same MIDI files when you send midi=true. The API reference at /docs/api lists the fields.
Can I use the MIDI in my own music?
Only if you have the rights to the song. Transcribing a part doesn't change who owns the composition.